A profitable business may reasonably choose not to grow. Another may need to fix service quality before pursuing rising demand. The answer does not come from one metric: it comes from comparing alternatives, safeguards, and the consequences of being wrong.
At the end of the Atlas's first layer, the question is no longer only “What do we measure?” It is which course of action deserves scarce time and capital without endangering the existing business. HM Treasury's Green Book 2026 sets out a structured assessment of objectives, options, costs, benefits, risks, and uncertainty, including business as usual [R01]. It was written for public-sector appraisal: we borrow the principle of comparing options and assumptions, not the government's social valuation rules as mandatory accounting for private small businesses.
1. Distinguish six genuinely different decisions
| Decision | Meaning | Illustrative signal, never an automatic rule |
|---|---|---|
| Optimize | Improve waste, experience, workflow, or cost in the current operation | A meaningful, addressable performance gap |
| Maintain and monitor | Sustain the offer with controls and periodic review | The operation works, but other investments have weak value or evidence is insufficient |
| Grow the existing operation | Serve more volume, hours, or locations | Repeatable demand, capacity, quality, and economics support it |
| Explore another offer | Pilot a variant, segment, channel, or model | Signals of a different need and a bounded way to test it |
| Reduce or pause | Restrict coverage, spending, or activity | Exposure exceeds resources or quality cannot be protected |
| Exit or discontinue | Stop an unsustainable or no-longer-useful activity | No defensible path to correct harm, noncompliance, or persistent losses |
Maintenance is not neglect: it requires monitoring, upkeep, compliance, and a review date. Growth is not a prize that automatically follows improvement. GOV.UK acknowledges that some services should be retired when users no longer need them or too few do to justify continued operation [R02]. That provides a management question, not a universal rule requiring a particular business to close.
2. Set gates that cannot be outweighed by a numerical score
Before ranking initiatives, separate mandatory gates from trade-offs. Mandatory gates include safety, applicable consumer and operating obligations, the ability to honor existing promises, and sufficient cash for due commitments. If these conditions fail, an otherwise attractive commercial score does not authorize execution.
Assess the remaining options through eight lenses: (1) repeatable demand and target segment; (2) incremental contribution; (3) available cash and payment timing; (4) service quality and user experience; (5) capacity and dependencies; (6) evidence and uncertainty; (7) supplier/customer concentration; and (8) alternatives and irreversibility. Use the business health assessment and consistent definitions from business KPIs.
3. Compare a business-as-usual case with actual changes
For every option specify what might happen if it is not implemented, the horizon, population, fixed and variable costs, upfront investment, working capital, risks, time needed, and economic result. Even business as usual can change with prices or demand; do not treat it as a perfectly static forecast.
Calculate incremental value under explicit assumptions, not revenue alone. Fifty additional sales can generate zero benefit if extra shifts, discounts, and returns absorb the contribution. Refer to how to evaluate improvement profitability and readiness to scale before committing to construction, leases, or new staff.
Separate reversible actions—adjusting a pilot shift—from investments that are difficult to recover—long leases, renovations, or specialized equipment. Account for opportunity cost: doing A may make B impossible because time or cash is limited, even if B is also attractive. The Green Book emphasizes that a numerical benefit-cost ratio alone is insufficient when options entail risks and constraints [R01].
4. Use a reasoned decision matrix, not a magic formula
For each option, compare intended outcome, observed evidence, forecast, capital required, incremental contribution, cash requirements, capacity, quality, compliance, uncertainty, reversibility, opportunity cost, and proposed decision. Weighted scoring can be used if its definitions are visible, but it must never hide the failure of a mandatory gate. An unexplained “8/10” is not analysis.
Provide at least three outcomes: Go when safeguards and sufficient learning permit action; Not yet when a specific check is missing; and No-Go when present conditions rule out the option. A Not yet recommendation must have an owner, trigger, and review date rather than becoming an indefinitely postponed idea.
5. Worked example: three choices for Repostería Horizonte
This is a fictional decision scenario, not established results from P7-13 or an audited business forecast. The bakery has limited improvement capacity and cannot deliver all three initiatives simultaneously. Assume S/1,500 is available for discretionary improvement commitments, with existing quality and service obligations protected.
| Option | Monthly economic assumptions | Initial outlay | Decisive risk | Preliminary decision |
|---|---|---|---|---|
| A. Reduce waste in the existing line | Avoid 12 discarded units at S/18 = S/216; extra labor S/60 ⇒ +S/156/month | S/300 | Do not increase stockouts or reduce quality | Optimization pilot |
| B. Extend opening hours | 55 extra orders × S/12 contribution = S/660; extra shift S/850 ⇒ −S/190/month | S/1,300 | Unproven demand and cash exposure | No-Go under current assumptions |
| C. Launch another cake | Possible monthly net contribution: −S/150 / +S/200 / +S/450, depending on demand and substitution | S/1,200 | Short pilot, cannibalization, uncertain repeat behavior | Not yet: investigate and retest |
Option A has a simple payback of 300/156 = 1.92 months if savings persist, no other costs arise, and shortages do not increase. It is an appraisal assumption, not a guaranteed gain. Option B loses S/190/month as modeled and would require at least 850/12 = 70.83, or 71 additional whole orders, just to cover the extra shift before recovering S/1,300. Demand for that volume has not been demonstrated.
For C, even the base case of S/200/month implies a six-month simple payback (1,200/200) if that contribution is repeatable—which remains unproven. With −S/150/month, the investment cannot pay for itself under the modeled conditions; S/450/month paints a different picture. That sensitivity calls for evidence rather than an irreversible commitment. Use exploring new business opportunities to design the next learning step.
Conditional decision: authorize a short, reversible test of A while recording waste, stockouts, contribution, and complaints; do not extend opening hours yet; keep C as a bounded learning exercise without authorizing a fixed investment. A serious safety incident or unmet legal obligation would take priority over any discretionary ranking.
6. Recognize misleading signals of success
Five common false positives are discount-driven volume without contribution, invoiced sales that have not been collected, favorable surveys from self-selected respondents, full capacity with late orders, and debt-financed growth that cannot meet its own payment schedule. For each number state the denominator, time horizon, and cost. A purchase by a new customer is not retention: consult customer cohorts and loyalty before appraising a program.
A before-and-after change coinciding with holidays, a new schedule, or a promotion does not establish causality. Operational improvement experiments can reduce alternative explanations where a suitable design is feasible. Even positive economic value may be unacceptable when safety, compliance, or customer treatment deteriorates.
7. Define triggers and evidence for revisiting the decision
Every decision should specify what will be observed, how often, what threshold triggers a review, who is accountable, and what happens in an adverse scenario. For A: enough comparable operating days, a waste criterion, stockouts, and labor cost. For B: verified additional demand, sufficient cash, and an on-time shift test. For C: a second realistically priced pilot, voluntary follow-up, net margin, and an updated substitution estimate.
Reassess when a material assumption changes, not just because the calendar says so. A different customer or business model may require returning to opportunity assessment, the Business Model Canvas, and rigorous validation design. Revisiting discovery when facts contradict a plan is not failure.
8. Deliverable: options matrix and decision record
Keep an options matrix with scenarios and non-negotiable gates and a decision record covering date, owner, accepted/rejected option, rationale, assumptions, residual risks, authorized budget, next evidence check, and rollback conditions. Attach baseline observations and separate them from modeled values. The editorial package includes both instruments in Spanish and English.
Frequent mistakes and limitations
A high score is not authorization; optimization is not expansion; revenue is not liquidity; interest is not paid demand; and uncertainty is not an excuse to ignore obligations. A business need not grow, nor should it preserve a harmful activity forever because money was spent on it previously. Legal, tax, and financial advice must be adapted to the actual jurisdiction and situation with qualified professionals and relevant authorities.
What you should have when finished
A verifiable choice among optimize, maintain, grow, explore, reduce, or stop, grounded in evidence and constraints, with an explicit condition for changing course. This completes the first layer of Pillar 7; the articles do not replace business judgment or ongoing monitoring.
References
- [R01] HM Treasury, The Green Book (2026): appraisal of costs, benefits and risks of options. https://www.gov.uk/government/publications/the-green-book--2/the-green-book-2026
- [R02] Government Digital Service (GOV.UK), How the live phase works. https://www.gov.uk/service-manual/agile-delivery/how-the-live-phase-works
- [R03] Government Digital Service (GOV.UK), Measuring the benefits of your service. https://www.gov.uk/service-manual/measuring-success/measuring-service-benefits